Safety culture is how a workplace really treats safety: what people do when no one is watching, what happens when someone raises a problem, and what leaders pay attention to. It cannot be changed by posters. It changes when leaders change how they respond and when the system stops asking people to work safely with the wrong tools and too little time.
This guide covers what culture is, the types of human error and what fixes each, the Swiss cheese model, human and organizational performance, a just culture, maturity levels, how to measure culture from surveys, observations, and reports, what leaders do, and the strengths and limits of behavior-based safety. A worked example uses a survey and a month of observations at an illustrative plant, with gaps between groups and the reasons behind at-risk work.
Survey and observation data in the example are illustrative.
Before You Start
What Safety Culture Is, and Is Not
Safety culture is the shared values, beliefs, and habits about safety in a workplace: what people do about safety when no one is watching, what they expect to happen when they raise a problem, and what leaders pay attention to. It is not a poster, a slogan, or an annual survey score.
| Term | Meaning | How it shows |
|---|---|---|
| Safety culture | Deep, shared assumptions about how things are done here | Whether a worker stops the line, whether a manager listens to bad news |
| Safety climate | People’s perceptions of safety at a point in time | What a survey measures |
| Safety behavior | What people do | What observations record |
| Safety system | The policies, procedures, and controls in place | What an audit checks |
You cannot change culture directly. You can change what leaders do, how the system responds to reports and mistakes, and what the work asks of people. Culture follows.
Human Error: A Starting Point, Not a Conclusion
People make errors. It is a normal part of human performance, and it happens most when the work is rushed, unclear, interrupted, or badly designed. Safe systems are designed with the assumption that errors will occur, so that errors do not lead to harm. Reason’s classification is useful for finding the right kind of fix.
| Type | What it is | Example | What helps |
|---|---|---|---|
| Slip | A correct plan carried out wrongly, an action error | Pressing the wrong button on a panel | Better design, labels, layout, forcing functions |
| Lapse | A memory failure, forgetting a step | Forgetting to release stored pressure | Checklists, interlocks, reminders at the point of work |
| Mistake | A wrong plan, from a wrong rule or a lack of knowledge | Believing the machine is isolated when it is not | Training, clear procedures, feedback, supervision |
| Violation, routine | A habitual shortcut that becomes normal | Bypassing a guard because everyone does | Find why the shortcut is needed; redesign the task |
| Violation, situational | A deviation under pressure, from the conditions | Skipping lockout to meet a deadline | Fix time, tools, staffing, and the signal from management |
| Violation, reckless | A conscious disregard of a known, substantial risk | Deliberately disabling a safety device for no reason | Accountability, which is a small minority of cases |
Human and Organizational Performance
Human and organizational performance (HOP) is a way of looking at safety that starts from how work really happens. A common summary has five principles, set out by Todd Conklin:
| Principle | What it means for leaders |
|---|---|
| Error is normal | Plan for it. Build defenses so that mistakes do not become injuries |
| Blame fixes nothing | Blame stops people from talking, and the learning stops with it |
| Context drives behavior | Change the work conditions and tools, not just the person |
| Learning is vital | Hear from the people who do the work, and treat near misses as gifts |
| How you respond matters | The first response to bad news decides whether you hear any more |
This links to the idea that safety is more than the absence of accidents. In Hollnagel’s terms, Safety-I studies failures and tries to prevent them. Safety-II also studies why things normally go right, and what people do to adapt to keep work safe, so that you can support that. A plant needs both: investigate what went wrong, and learn from the many times that people made it go right.
A Just Culture
A just culture makes a clear line between honest mistakes, which the system should learn from, and conduct that deserves accountability. People who fear punishment for honest errors will hide them. People who see no accountability for reckless acts will lose trust in the system.
| Type of behavior | Typical response |
|---|---|
| Human error: an unintended slip or lapse | Console the person; look at the system; fix the design |
| At-risk behavior: a shortcut where the risk is not recognized or is believed to be justified | Coach the person; find out why the shortcut is attractive; remove the incentive |
| Reckless behavior: a conscious disregard of a substantial and unjustifiable risk | Hold the person accountable, within the organization’s policy and the law |
A useful check is the substitution test: would two or three other people with similar training and in the same conditions have done the same? If so, the cause is in the system and not in the person.
These categories follow the “just culture” model of David Marx. Use them as a guide and not a formula, and have your HR and legal advisers review your own policy before using it for discipline.
Culture Maturity
| Level | How it responds to bad news | Typical signs |
|---|---|---|
| Pathological | Hides it or punishes the messenger | Blame, cover-up, safety as a cost |
| Reactive | Responds after an accident | Actions after incidents, then they fade |
| Calculative | Manages it with systems and numbers | Many procedures and audits; the paper looks good; little ownership on the floor |
| Proactive | Looks for it and acts early | Reports rise, leaders walk the floor, near misses are used |
| Generative | Welcomes it as information; integrates it | Safety is built into the work, and everyone owns it |
The levels are a way to start a conversation about where you are, not a scorecard. Most organizations are a mix, and different areas differ.
Measuring Culture
Use more than one source. A perception survey shows what people think. Observations show what people do and why. The system shows how the organization responds.
| Source | What it tells you | Cautions |
|---|---|---|
| Perception survey | Climate: how people rate management commitment, reporting, involvement, and resources | Anonymous; compare groups, not individuals; keep it short; act on the result |
| Observations and gemba walks | What people actually do, and what makes safe work hard | Do them as conversations, record barriers, not just behaviors |
| Near-miss and hazard reports | Trust in the reporting system | A rise is usually good |
| Leading indicators | Whether leaders and the system act | Action closure, walk completion, training |
| Interviews and focus groups | The reasons behind the numbers | Skilled neutral facilitation |
- Look at the gap between groups. The distance between managers’ and operators’ answers is often more telling than the average.
- Say what you will do before you survey. A survey that is followed by silence harms trust.
- Use small-sample care. A group of twelve is a small sample, and a few answers can move the percentage. Show the uncertainty and avoid reading too much into one question.
Worked Example: A Culture Check at Riverside Plant
Riverside Plant, an illustrative 140-employee metal fabrication and assembly plant, ran an anonymous eight-question safety climate survey and a month of structured gemba observations. 96 of 140 employees responded (69%): 12 managers, 14 supervisors, and 70 operators. All data are illustrative.
| Dimension | Managers (n=12) | Supervisors (n=14) | Operators (n=70), 95% interval | Gap (points) | p (managers vs operators) |
|---|---|---|---|---|---|
| Management commitment | 92% | 86% | 71% (60 to 81) | 20 | 0.173 |
| Worker involvement | 92% | 79% | 59% (47 to 69) | 33 | 0.048 |
| Reporting without fear | 92% | 71% | 49% (37 to 60) | 43 | 0.010 |
| Stop-work authority | 100% | 86% | 66% (54 to 76) | 34 | 0.015 |
| Training and competence | 83% | 79% | 74% (63 to 83) | 9 | 0.721 |
| Communication and feedback | 83% | 71% | 51% (40 to 63) | 32 | 0.058 |
| Resources and time | 83% | 64% | 44% (33 to 56) | 39 | 0.026 |
| Peer support | 83% | 86% | 79% (68 to 87) | 5 | 1.000 |
Reading it. On peer support and training, all three groups agree. On reporting without fear, stop-work authority, and resources and time, managers see a strong system and operators do not: fewer than half the operators say they can report without fear, and fewer than half say there is enough time and resources. The p-values come from an exact test for each question. With eight questions and small groups, treat them as signals and not proof: after allowing for eight comparisons, none would pass a strict test, but the pattern across questions is the point.
The observations. Over the same month, trained observers made 60 short observations on the floor and recorded 23 at-risk situations. For each, they asked the worker why. The reasons are in the chart below. Fewer than one in seven (3) were habit or drift. 12 of the 23 came from tools that were not suited to the task or from time pressure.
Use the Safety Gemba and Observation Checklist to run observations and see the barrier chart and the share of safe observations with an interval.
What Leaders Do
- Be visible, and listen first. Walk the floor, ask open questions, and ask what makes safe work hard.
- Respond well to bad news. Thank the person, ask questions, and fix it. The first reaction decides whether you hear more.
- Back stop-work. Publicly support the first person who stops a job, and never make them regret it.
- Give resources and time. A requirement to work safely, without time and tools to do it, is a mixed message.
- Hold themselves to the same rules. Wear the PPE, follow the lockout, and take part in the audits.
- Close the loop. Tell people what happened to the problems they raised, and show the before and after.
- Measure what you want more of, such as reports, closed actions, and walks, and not only injuries.
Behavior-Based Safety: Strengths and Limits
Behavior-based safety (BBS) uses observations of what people do, with feedback and coaching, to build safer habits. Used well, it brings supervisors and workers into frequent conversations and gives data on practice. It is described in the Behavior-Based Safety entry.
| Used well | Used badly |
|---|---|
| Observations are conversations that find out why | Observations are checklists that catch people |
| Barriers in the system are recorded and fixed | The worker is the problem, so the system stays as it is |
| Reports are valued | Counts of observations or “safe behavior” scores are used to rank or reward teams |
| Used alongside hazard control and engineering | Used in place of hazard control |
The main criticism is that BBS can turn into worker blame and divert attention from hazards that should be controlled by design. Keep it a tool for learning about the work, with the hierarchy of controls first.
Common Mistakes
Culture as Posters
Slogans and signs with no change in how the system responds.
Blaming the Last Person
“Human error” closes investigations.
Survey and Silence
Asking for opinions and doing nothing with them.
Rewarding Zero Injuries
Bonuses that teach people to hide injuries.
Observing to Catch
Behavior checklists that put people on the defensive.
Leaders Exempt
Rules that apply to the floor and not the office.
Self-Assessment Questions
- What happens, in practice, to a person who reports a mistake or a near miss?
- Do operators and managers answer culture questions the same way? If not, where is the gap?
- Do investigations stop at “human error,” or ask why it made sense at the time?
- Do leaders spend time on the floor, and do they leave knowing something that they did not know before?
- What do our observations say are the reasons for at-risk work?
Safety Culture and Human Performance: Frequently Asked Questions
What is safety culture?
The shared values, beliefs, and habits about safety in a workplace: what people do when no one is watching, how they expect the organization to respond when they raise a problem, and what leaders pay attention to. It is distinct from the safety system (policies and controls) and from safety climate (a measure of perceptions at one time).
How do you measure safety culture?
Use several sources: an anonymous perception survey compared across groups, structured observations and gemba conversations that record barriers, trends in near-miss and hazard reporting, leading indicators such as action closure, and interviews. Compare managers and operators, because the gap is often more informative than the average.
What is a just culture?
A culture that separates honest errors, which the system should learn from, from at-risk and reckless behavior, which call for coaching and accountability. The aim is to keep people reporting and learning while holding to account the conduct that deserves it. The substitution test helps: would others in the same conditions have acted the same way?
What is the Swiss cheese model?
James Reason’s picture of layered defenses, each with holes. A hazard reaches a person only when the holes in several layers line up. It shows why a single mistake rarely causes harm by itself, and why fixing the latent conditions in the system is usually more effective than blaming the last person.
Is behavior-based safety effective?
It can be, when observations are conversations that find out why people work as they do, barriers are fixed, and it is used alongside control of hazards. It is criticized when it becomes worker blame, ranks teams by scores, or takes the place of engineering controls. Keep the hierarchy of controls first.
Sources and Further Reading
- Reason, J., Human Error (Cambridge University Press, 1990) and Managing the Risks of Organizational Accidents (Ashgate, 1997).
- Dekker, S., The Field Guide to Understanding ‘Human Error’ (3rd ed., CRC Press, 2014).
- Hollnagel, E., Safety-I and Safety-II: The Past and Future of Safety Management (Ashgate, 2014).
- Conklin, T., The 5 Principles of Human Performance (PreAccident Investigation Media, 2019).
- Westrum, R., “A typology of organisational cultures,” Quality and Safety in Health Care, 13 (Suppl 2), 2004; Hudson, P., “Safety management and safety culture: the long, hard and winding road,” 2001.
- Marx, D., Patient Safety and the “Just Culture”: A Primer for Health Care Executives (Columbia University, 2001).
- ISO 45001:2018, clause 5 on leadership and worker participation (check current edition); UK Health and Safety Executive, Managing for Health and Safety (HSG65).